Nearly all AI industry revenue comes from 30 million programmers; 1.5 billion knowledge workers are just beginning
This wave of AI revenue is almost entirely supported by 30 million programmers as heavy users; enterprise penetration is below 5%; once it spreads to 1.5 billion knowledge workers, the market could grow dozens of times larger.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Nearly all enterprise AI revenue depends on just 30 million programmers
While the world has 1.5 billion knowledge workers, the vast majority of enterprise AI revenue comes from a single group of just 30 million—programmers. Even within that 30 million, there's an extreme power law: roughly 1 million at the top drive most of the spending. David George sees this as proof the opportunity is huge: penetration today is still microscopic. Once it spreads to a broader base of knowledge workers, the market will dwarf today's size.
— David GeorgeEnterprise penetration below 5% while supply is already booked through 2028
Except for coding and writing, AI's spread through enterprises remains early-stage. Law is an exception—roughly 12 months behind coding, but overall penetration into the knowledge-work market is below 5%. Supply-side pressure is tighter still: new data-center capacity is booked through 2028, because every link in the supply chain is a bottleneck.
— David GeorgeLegal AI has become a forced standard, but giants won't build it themselves
Legal AI, exemplified by Harvey, has moved past the ‘will it hallucinate?’ phase. Now end clients are demanding their law firms use it—both because the product works and for cost reasons. Kirkland's announcement of $500 million invested in building its own tech stack was the signal moment. But David George argues frontier labs won't build vertical products like legal themselves, because deployment demands extensive hands-on product work and field sales, which rank maybe 6th to 15th on their priority list.
— David GeorgeThe last wave created $25 trillion in value; this one will be larger
Mobile internet, social, e-commerce, SaaS, and cloud—that generation of tech waves created roughly $25 trillion in market value combined. Now happening simultaneously are AI, autonomous-driving robotics, U.S. manufacturing modernization, and AI applied to biohealth. Layered together, these waves will inevitably dwarf $25 trillion, and in most categories the major players haven't even entered the field yet.
— David GeorgeWaymo is 10–14× safer than human drivers; cost is the real constraint
By mile-based safety data, Waymo is already 10–14× safer than a human driver. David George argues the real consumer inflection point isn't technology but cost: driving yourself costs roughly $0.80 per mile, Uber or Lyft over $2. Once self-driving hits sub-private-car per-mile cost, demand will explode. Today, Waymo has fewer than 10,000 vehicles across the whole U.S.
— David GeorgeAI breaks the VC power law: massive capital actually improves the product
Historically, private-market returns split evenly between seed-to-B and C onward; that ratio is shifting toward later stage. More critically, this AI wave's power law is more extreme than before: throwing hundreds of billions into a SaaS company just breaks the team (Vision Fund's lesson). But hundreds of billions spent on model training actually does make the model better—a returns-to-scale dynamic unique to AI.
— David GeorgeProduct cycle scores 9–10; funding cycle only 6
David George scores ‘product cycle’ and ‘funding cycle’ on a 1–10 scale. In 2021, product cycle was 1—mobile had run out of runway. In 2010, it was 8. Now, with AI, autonomous driving, and U.S. manufacturing modernization happening at once, product cycle is 9–10. Funding cycle (valuation cheapness) is only 6—good, but not cheap. The two almost never hit high simultaneously.
— David GeorgeVCs must own the narrative, and Ben Horowitz made that clear immediately
Before joining a16z, David George told Ben Horowitz he didn't need to do marketing. Ben told him he was wrong on the spot. Narrative matters because it directly shapes valuation, fundraising ability, and hiring and retention. Palantir proves it: most shareholders couldn't explain ontology, yet the valuation premium and ‘trusted AI practitioner’ positioning genuinely drove more commercial deals.
— David GeorgeIn their own words · checked verbatim
I am very, very closely aligned with Eric on this point that it's all going to work. I actually like the way that he framed it up. I framed it up slightly differently, which is just like the answer to all this is and.
David George1:07
There's 1.5 billion knowledge workers. But like basically all of this, which is like adding more revenue than the best businesses that have ever been created, is on the back of a small group of people.
David George3:13
there was a major validation event, which was when Kirkland came out and said they were going to spend 500 million bucks to build their own technology stack.
David George16:51
So like it is barbaric to not actually allow for these things to diffuse into your economy.
David George29:12
We did this analysis a long time ago that basically showed half of private market returns get generated between the seed and the B. And then half of returns get generated from the C plus.
David George36:13
And he's like, well, what's your marketing strategy? And I was like, we don't really need a marketing stretch. It's not really a thing in growth. And he just like looked at me, and said like, that's the dumbest fucking idea I've ever heard.
David George44:17
I never short a Messianic founder, and I never short a product that people love.
David George47:23
Figures
| OpenAI + Anthropic + xAI + Cursor combined annualized revenue | ~$120 billion | 2:13 |
| Programmers supporting enterprise AI revenue | 30 million | 3:13 |
| Global knowledge-worker base | 1.5 billion | 3:13 |
| AI penetration in B2B economy | Below 5% | 4:14 |
| Top three frontier labs combined revenue | ~$125 billion | 11:41 |
| OpenAI + Anthropic cumulative fundraising | $350 billion | 11:41 |
| Waymo safety margin vs. human drivers | 10–14× | 29:12 |
| Market value created by previous tech cycle | $25 trillion | 27:10 |
Glossary
- skeuomorphic
- Using new technology to mimic the interface of an old product, rather than exploiting what makes the new technology distinct.
- blast radius
- The adjacent product categories a frontier AI lab considers worth building in-house rather than leaving to partners.
- model busters
- David George's term for companies like Databricks that repeatedly reinvent themselves via AI and see their valuations chronically underestimated.
- harness
- The product-layer code that wraps a model and determines how it interacts with users and its environment.
How to listen
Founders tracking enterprise AI adoption rates, VCs focused on late-stage growth investing, and anyone trying to understand the commercial inflection point for autonomous driving.
From 50:32 onward, the segment on a16z's own scale and market-share ambitions has limited value for non-VC professionals.